A0488
Title: LIMOS--LightGBM interval Merton one-period-portfolio selection
Authors: Liang-Ching Lin - National Cheng Kung University (Taiwan) [presenting]
Sz-Wei Charng - National Cheng Kung University (Taiwan)
Abstract: The modern portfolio theory can assist us in allocating wealth to risky and risk-free assets reasonably by using some statistical methods. The aim is to focus on evolving Merton's portfolio problem. Instead of the conventional parameter estimations based on only the closing prices, the opening, high, low, and closing prices are included to enlarge the database as much as possible to make the parameter estimations much more accurate. Furthermore, a weighted arithmetic mean of estimations obtained is considered from different lengths of training datasets to stabilize the estimators in which the weights are evaluated by using the least-squares method. In addition, the LightGBM is used to predict the transaction directions, and not only the prices as tradition and also many statistics are included to be the features. In real data analysis, it is demonstrated the usefulness of combining the aforementioned methods by showing the portfolio profits of selecting ten stocks in 2018 and 2019. The results particularly show the superiority of the proposed strategy over the conventional method: the profits are almost positive and have around 32\% to 72\% annually.